Anthropic had quite a week: cleaner plan names, fewer limits for power users, and a malware scare hiding inside the very tool developers trust most.
The irony? As Claude gets more capable and agentic, it also becomes a bigger target. A poisoned extension planting SSH backdoors on 5,000 dev machines is a reminder that the supply chain is now the attack surface.
We're building faster than we're securing.
|
|
In Partnership with
|
·
|
|
|
|
|
|
Today's Author
Lior Alexander. Founder of AlphaSignal and former ML Engineer. Previously built ML systems at Iguazio, Guesty, Enphase, Mila.
|
|
Top News
|
|
|
|
Anthropic renames Claude plans and drops 5-hour limit for verified Ultra subscribers
|
|
31,853 Likes
|
|
Claude just got a naming refresh. The Max 5x plan is now simply called Max ($100/month), and Max 20x is now called Ultra ($200/month). Same price, cleaner names. Nothing changes for existing subscribers. The bigger news? The 5-hour usage limit is gone for verified Ultra accounts. That cap used to cut off heavy sessions mid-flow, which was frustrating if you were deep in a long coding task or running background jobs. Now you can run longer without hitting a wall. A few things to know: - Newer accounts keep a temporary limit until verified, to prevent abuse. You'll get an email when it's lifted.
- Weekly limits are still in place, unchanged.
- If you're already verified, the removal is rolling out to you now.
Bottom line: longer uninterrupted sessions means you can tackle bigger tasks without getting cut off at a bad moment.
|
|
|
|
Presented by Tiger Data
|
|
100M Vectors, 6ms Latency, No New Database
|
|
Vector search in Postgres usually hits a wall: HNSW indexes get huge and slow to build, pushing teams to bolt on a separate vector database. pg_vectorsearch removes that tradeoff so you can stay in Postgres.
Its new PRISM index, tested at 100M vectors:
• 6ms p50 latency at 95% recall, on 4x less memory • Index builds in under 2 hours, not 42 • Works on your existing pgvector columns, no rewrite
Open source and in alpha.
|
|
|
|
Top Repo
|
|
|
|
Developer rebuilds Apple's subvocal feature for AirPods Pro in 9 minutes with one prompt
|
|
9,792 Likes
|
|
A startup called Subvocal raised $1.3M to build glasses that turn silent mouthing into text at 200+ words per minute. No keyboard. No talking out loud. Just mouth words and they appear as text. Then Joe Blau replied with a gut punch: he rebuilt the same idea in 9 minutes with one prompt, for AirPods you already own, and put it on GitHub for free. Here is how his app, subvox, actually works: - Your AirPods have motion sensors for spatial audio. Subvox reads that head-motion data when you silently mouth words.
- You train it by mouthing a word a few times. It stores those as templates, then matches future mouthing against them.
- No AI model involved. Just pattern matching on motion signals.
- To run it: brew install xcodegen, then xcodegen generate, then open in Xcode. Requires macOS 14 and AirPods Pro.
The bigger story: a funded startup took months to demo this. One prompt rebuilt it. Software ideas alone are not a moat anymore.
|
|
|
|
Presented by Pave
|
|
How will the team describe your next project?
|
|
Give them something to talk about. Pave builds production-ready apps that connect to 350+ tools and automate everyone’s manual workflows. Go ahead, be the hero. Build your custom app in minutes.
|
|
|
|
Top Repo
|
|
|
|
Vercel Labs ships open-source tool letting AI agents control your whole desktop
|
|
9,473 Likes
|
|
Most AI agents that control your computer work by taking screenshots and guessing where buttons are. Slow, expensive, and fragile. agent-desktop from Vercel Labs fixes that. Instead of screenshots, it reads your OS's accessibility tree (think of it as a structured map of every button, input, and window on screen) and gives each element a stable ID called a ref. Your agent clicks the ref, not a pixel. No vision model needed. Here's what you can do with it: - Let agents open apps, toggle settings, or inspect logs while you keep coding
- Run multiple agents in parallel, each with its own cursor, on the same machine
- Automate end-to-end desktop app testing without pasting screenshots into chat
- Works with Claude Code, Codex, and any shell-capable agent
Install it: npm install -g agent-desktop. Free, open-source, Apache 2.0. Runs on macOS, Windows, and Linux.
|
|
|
At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 300,000 developers stay informed and ahead. We're passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are.
Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We're based in the United States. By using our site, you agree to be governed by U.S. laws.
|
|
|
Looking to promote your company, product, service, or event to 300,000+ AI developers?
|
|
|
|
|